Advanced Micro Devices Corporation

Shomit N Das

Matthew J Tomei

Bradford Michael Beckmann

Seyedmohammad SeyedzadehDelcheh

John Kalamatianos

Greg Sadowski

David Allen Wood

Wei T Huang

Anthony T Gutierrez

Karthik Rao

Vedula Venkata Srikant Bharadwaj

Shrikanth Ganapathy

Vignesh Adhinarayanan

Kishore Punniyamurthy

Xudong An

Xianwei Zhang

Steven E Raasch

Wayne Burleson

Abhinav Vishnu

Srikant Bharadwaj

Philip B Bedoukian

Seyed Mohammad Seyedzadehdelcheh

Philip Bedoukian

Systems, methods, devices, and computer-implemented instructions for processor power management implemented in a compiler. In some implementations, a characteristic of code is determined. An instruction based on the determined characteristic is inserted into the code. The code and inserted instruction are compiled to generate compiled code. The compiled code is output.

Compiler directed fine grained power management

A processor determines losses of samples within an input volume that is provided to a neural network during a first epoch, groups the samples into subsets based on losses, and assigns the subsets to operands in the neural network that represent the samples at different precisions. Each subset is associated with a different precision. The processor then processes the subsets in the neural network at the different precisions during the first epoch. In some cases, the samples in the subsets are used in a forward pass and a backward pass through the neural network. A memory configured to store information representing the samples in the subsets at the different precisions. In some cases, the processor stores information representing model parameters of the neural network in the memory at the different precisions of the subsets of the corresponding samples.

Dynamic precision scaling at epoch granularity in neural networks

A method includes, in response to receiving an instruction to perform a first operation on first data stored in a memory device, obtaining first compression metadata from the memory device based on an address for the first data, and reducing a number of operations in a set of operations based on the first operation and one or more matching addresses, the one or more matching addresses corresponding to second compression metadata matching the first compression metadata.

Compression metadata assisted computation

Systems, apparatuses, and methods for implementing a multi-tiered approach to cache compression are disclosed. A cache includes a cache controller, light compressor, and heavy compressor. The decision on which compressor to use for compressing cache lines is made based on certain resource availability such as cache capacity or memory bandwidth. This allows the cache to opportunistically use complex algorithms for compression while limiting the adverse effects of high decompression latency on system performance. To address the above issue, the proposed design takes advantage of the heavy compressors for effectively reducing memory bandwidth in high bandwidth memory (HBM) interfaces as long as they do not sacrifice system performance. Accordingly, the cache combines light and heavy compressors with a decision-making unit to achieve reduced off-chip memory traffic without sacrificing system performance.

Resource-aware compression

Systems, methods, and devices for performing pattern-based cache block compression and decompression. An uncompressed cache block is input to the compressor. Byte values are identified within the uncompressed cache block. A cache block pattern is searched for in a set of cache block patterns based on the byte values. A compressed cache block is output based on the byte values and the cache block pattern. A compressed cache block is input to the decompressor. A cache block pattern is identified based on metadata of the cache block. The cache block pattern is applied to a byte dictionary of the cache block. An uncompressed cache block is output based on the cache block pattern and the byte dictionary. A subset of cache block patterns is determined from a training cache trace based on a set of compressed sizes and a target number of patterns for each size.

Pattern-based cache block compression

An asynchronous pipeline includes a first stage and one or more second stages. A controller provides control signals to the first stage to indicate a modification to an operating speed of the first stage. The modification is determined based on a comparison of a completion status of the first stage to one or more completion statuses of the one or more second stages. In some cases, the controller provides control signals indicating modifications to an operating voltage applied to the first stage and a drive strength of a buffer in the first stage. Modules can be used to determine the completion statuses of the first stage and the one or more second stages based on the monitored output signals generated by the stages, output signals from replica critical paths associated with the stages, or a lookup table that indicates estimated completion times.

Controlling the operating speed of stages of an asynchronous pipeline

In some embodiments, a memory controller in a processor includes a base value cache, a compressor, and a metadata cache. The compressor is coupled to the base value cache and the metadata cache. The compressor compresses a data block using at least a base value and delta values. The compressor determines whether the size of the data block exceeds a data block threshold value. Based on the determination of whether the size of the compressed data block generated by the compressor exceeds the data block threshold value, the memory controller transfers only a set of the compressed delta values to memory for storage. A decompressor located in the lower level cache of the processor decompresses the compressed data block using the base value stored in the base value cache, metadata stored in the metadata cache and the delta values stored in memory.

Data compression system using base values and methods thereof

In some examples, thermal aware optimization logic determines a characteristic (e.g., a workload or type) of a wavefront (e.g., multiple threads). For example, the characteristic indicates whether the wavefront is compute intensive, memory intensive, mixed, and/or another type of wavefront. The thermal aware optimization logic determines temperature information for one or more compute units (CUs) in one or more processing cores. The temperature information includes predictive thermal information indicating expected temperatures corresponding to the one or more CUs and historical thermal information indicating current or past thermal temperatures of at least a portion of a graphics processing unit (GPU). The logic selects the one or more compute units to process the plurality of threads based on the determined characteristic and the temperature information. The logic provides instructions to the selected subset of the plurality of CUs to execute the wavefront.

Apparatus and method for providing workload distribution of threads among multiple compute units

A processing device is provided which includes memory comprising data cache memory configured to store compressed data and metadata cache memory configured to store metadata, each portion of metadata comprising an encoding used to compress a portion of data. The processing device also includes at least one processor configured to compress portions of data and select, based on one or more utility level metrics, portions of metadata to be stored in the metadata cache memory. The at least one processor is also configured to store, in the metadata cache memory, the portions of metadata selected to be stored in the metadata cache memory, store, in the data cache memory, each portion of compressed data having a selected portion of corresponding metadata stored in the metadata cache memory. Each portion of compressed data, having the selected portion of corresponding metadata stored in the metadata cache memory, is decompressed.

Device and method for data compression using a metadata cache

Entropy agnostic data encoding includes: receiving, by an encoder, input data including a bit string; generating a plurality of candidate codewords, including encoding the input data bit string with a plurality of binary vectors, wherein the plurality of binary vectors includes a set of deterministic biased binary vectors and a set of random binary vectors; selecting, in dependence upon a predefined criteria, one of the plurality of candidate codewords; and transmitting the selected candidate codeword to a decoder.

Entropy agnostic data encoding and decoding

A processing system scales power to memory and memory channels based on identifying causes of stalls of threads of a wavefront. If the cause is other than an outstanding memory request, the processing system throttles power to the memory to save power. If the stall is due to memory stalls for a subset of the memory channels servicing memory access requests for threads of a wavefront, the processing system adjusts power of the memory channels servicing memory access request for the wavefront based on the subset. By boosting power to the subset of channels, the processing system enables the wavefront to complete processing more quickly, resulting in increased processing speed. Conversely, by throttling power to the remainder of channels, the processing system saves power without affecting processing speed.

Growing community of inventors

Austin, TX, United States of America

Shomit N Das